Net load (NL) refers to the difference between gross demand and variable renewable energy (RE) generation. As the integration of RE continues to grow, the variability and uncertainty of power systems increase, creating significant challenges for power system operators due to the intermittent behavior of RE sources. NL forecasting (NLF) has become crucial to maintaining the reliable and secure operation of power systems. This paper critically reviews the state-of-the-art NLF models in RE-integrated power systems, with a particular focus on emerging techniques such as Transformer models and generative artificial intelligence (Gen-AI) models. This paper systematically explores and analyses various data preprocessing, feature engineering (feature selection and extraction), and hyperparameter tuning methods as well as forecasting engines used in NLF. Furthermore, this paper categorizes the NLF models with corresponding strengths and challenges. In addition, an assessment of the interpretability, scalability, and applicability of NLF models is presented, highlighting their performance under varying levels of RE penetration. This paper also outlines challenges for NLF, e.g., the quality of data and the spatial and temporal scales. Finally, the future research directions are outlined based on the identified gaps in NLF.